README.md

---
license: cc-by-nc-sa-4.0
task_categories:
- tabular-classification
- tabular-regression
- feature-extraction
- image-feature-extraction
- text-classification
language:
- en
tags:
- agentic-ai
- synthetic-data
- data-science
- benchmark
- human-AI-collaboration
pretty_name: AgentDS-Insurance
---

# 🛡️ AgentDS-Insurance

This dataset is part of the **AgentDS Benchmark** — a multi-domain benchmark for evaluating human-AI collaboration in real-world, domain-specific data science.

**AgentDS-Insurance** includes structured policyholder and claims data for 3 challenges:

- Claims complexity prediction  
- Next-year loss estimation for pricing  
- Fraud detection from claim patterns  

👉 Files are organized in the `Insurance/` folder and reused across challenges.  
Refer to the included `description.md` for:
- File usage and challenge mappings  
- Task descriptions and data schema notes  
- Submission format expectations  

---



## 📑 Citation

Please cite AgentDS if you use it in research:

```bibtex
@misc{luo2026agentds,
  author       = {An Luo and Jin Du and Xun Xian and Robert Specht and Fangqiao Tian and Ganghua Wang and Xuan Bi and Charles Fleming and Ashish Kundu and Jayanth Srinivasa and Mingyi Hong and Rui Zhang and Tianxi Li and Galin Jones and Jie Ding},
  title        = {AgentDS Technical Report: Benchmarking the Future of Human-AI Collaboration in Domain-Specific Data Science},
  year         = {2026},
  eprint       = {2603.19005},
  archivePrefix= {arXiv},
  primaryClass = {cs.LG},
  note         = {arXiv:2603.19005},
  url          = {https://arxiv.org/abs/2603.19005}
}
```

See arxiv.org/abs/2512.20959 for an example of how we create data for a challenge like this.

---

📖 **More info & challenge details**: https://agentds.org/domains  
🔐 **Get your API key**: https://agentds.org/dashboard  
🧠 **Submit predictions via SDK**: `pip install agentds-bench` (see main AgentDS README for usage)